For "almost all" sufficiently large $N,$ satisfying necessary congruence conditions and $k\geq 2$, we show that there is an { \bf asymptotic formula} for the number of solutions of the equation \begin{ align*} \begin{ split} &N=p_{ 1}^{ k}+p_{ 2}^{ k}+\cdots+p_{ s}^{ k}, \\ &\left|p_{ i}-( N/s)^{ 1/k}\right|\leq (N/s)^{ \theta/k},\ (1\leq i\leq s) \end{ split} \end{ align*} with \begin{ align*} s\geq \frac{ k(k+1)}{ 2}+1\ \ \textup{ and}\ \ \theta\geq { \bf 2/3}+\varepsilon. \end{ align*} This enlarges the effective range of $s$ for which can be obtained by the method of M\"{ a}tomaki and Xuancheng Shao \cite{ MS}. [Discorrelation between primes in short intervals and polynomial phase, Int. Math. Res. Not. IMRN 2021, no. 16, 12330-12355.] The idea is to avoid using the exponential sums (1.2) and Vinogradov mean value theorems in Lemma 2.4 simultaneously. And the main new ingredient is from Kumchev and Liu \cite{ KL} (see Lemma 2.2).
Categories: math.NT